add length for faster train

main.py

@@ -61,6 +61,8 @@ def prepare_dataset(batch, processor):
 
     batch["input_values"] = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0]).input_values
 
+    batch["length"] = [len(item) for item in batch["input_values"]]
+
     with processor.as_target_processor():
         batch["labels"] = processor(batch["target_text"]).input_ids
     return batch